18th International Specialty Conference on Cold-Formed Steel Structures

October 26-27, 2006, Orlando, Florida

Buckling analysis of cold-formed steel members using CUFSM:

conventional and constrained finite strip methods
B.W. Schafer1 and S. Ádány2

Abstract

The objective of this paper is to provide technical background and illustrative

examples for stability analysis of cold-formed steel members using the
conventional and constrained finite strip methods as implemented in the open
source program CUFSM. Numerical stability analysis combined with an
accurate identification of the buckling modes is an enabling first step in the
implementation of new design methods such as the Direct Strength Method. In
this paper conventional finite strip method analysis identical to that employed in
CUFSM is derived from first principles. The methodology closely mirrors that
of standard matrix methods for structural analysis, widely used by engineers,
and can be readily programmed by the interested reader. An example of the
stability solution for an industry standard lipped channel is provided. Recently,
CUFSM has been extended to include application of the constrained finite strip
method. Using formal mechanical definitions of the buckling classes: global,
distortional, local, and other deformations, the constrained finite strip method
can provide both modal decomposition and modal identification to a
conventional finite strip solution. Modal decomposition allows the conventional
finite strip solution to be focused on any buckling class (e.g., global, distortional,
or local only), resulting in problems of reduced size and definitive solutions for
the buckling modes in isolation, as demonstrated for an example section. Modal
identification allows the results of a conventional finite strip solution to be
judged with regard to the participation of the buckling classes; and thus provide
a measure of buckling mode interaction. The conventional finite strip method
combined with the constrained finite strip method provides a powerful tool for
understanding cross-section stability in cold-formed steel members.

1
Associate Professor, Johns Hopkins University, United States,
schafer@jhu.edu
2
Associate Professor, Budapest University of Technology and Economics,
Hungary, sadany@epito.bme.hu
Introduction

Cold-formed steel members are thin, light, and efficient. However, this
efficiency comes with complication: engineers must consider buckling of the
thin walls of the cross-section in addition to global (e.g., flexural or lateral-
torsional) buckling of the member. Classical hand solutions to these instabilities
become unduly cumbersome for more complex buckling modes, such as
distortional buckling; and may ignore critical mechanical features, such as inter-
element equilibrium and compatibility. To remedy this, the engineer may turn to
numerical solutions such as the finite strip method (FSM).

Conventional FSM provides a means to examine all the possible instabilities in a

cold-formed steel member under longitudinal stresses (axial, bending, or
combinations thereof). CUFSM is an open source FSM program, freely
available from the first author of this paper, that provides engineers with this
capability (www.ce.jhu.edu/bschafer/cufsm). The basic framework of the FSM
stability solution will be familiar to anyone who has studied matrix structural
analysis. In this paper the underlying elastic and geometric stiffness matrices are
derived and shown in closed-form so that users can become more familiar with
this tool and more comfortable with employing FSM solutions in design. In
addition, an example of a conventional FSM solution is provided. New design
methods such as the Direct Strength Method become highly efficient for the
engineer when FSM stability solutions are employed.

Recently, extensions to the conventional FSM solution have been explored,

namely, the constrained finite strip method, or cFSM. A cFSM solution provides
a means for (1) stability solutions to be focused only a given buckling mode
(modal decomposition), or (2) a conventional FSM stability solution may be
classified into the different fundamental buckling modes (modal identification).
Modal decomposition is accomplished by forming a series of constraint
equations which describe a particular buckling class, i.e., local, distortional, or
global buckling. A conventional FSM solution is then constrained to the selected
buckling class and stability analysis performed. Since the number of degrees of
freedom (DOF) within a class are much less than the full model, modal
decomposition also results in significant model reduction. In this paper the
criteria used to define the classes are provided, and implementation of the
constraints is demonstrated. Modal identification employs the developed
constraint equations as a means to transform the solution basis from the nodal
DOF to modal DOF associated with the classes. This allows a buckling mode to
be identified in terms of its contribution from the different classes. Formal
definition and an example of modal identification are provided. The
implementation of cFSM discussed herein is completed in CUFSM.
Conventional Finite Strip Method

In the finite strip method (FSM) a thin-walled member, such as the lipped
channel of Figure 1, is discretized into longitudinal strips. The advantage of
FSM over other methods, such as the finite element method which applies
discretization in both the longitudinal and transverse direction, is dependent on a
judicious choice of the shape function for the longitudinal displacement field. In
Figure 1 a single strip is highlighted, along with the degrees of freedom (DOF)
for the strip, the dimensions of the strip, and applied edge tractions (loads).
b
v1
v2

θ1 θ2
a y
u1 u2
w1 w2
x

v1 z
v2
T1=f1t T2=f2t

Figure 1 Finite strip discretization, strip DOF, dimensions, and applied edge tractions

Elastic stiffness matrices

The vector of general displacement fields: u = [u v w]T is approximated by the

displacement at the nodes, d, and selected shape functions, N, via
u = Nd = [Nuv Nw] [duvT | dwθΤ]T
where the nodal displacements in local coordinates are shown in Figure 1 and
may be explicitly written as d = [u1 v1 u2 v2 | w1 θ1 w2 θ2]T or in partitioned form
d = [duvT | dwθΤ]T. For the in-plane, or membrane, displacements u and v linear
shape functions are employed in the transverse direction, and in the longitudinal
direction u employs a sin function and v a cos function for displacement:
 x   x   u   mπy   x   x   v1   mπy 
u = 1 −     1  sin  ,v= 1 −    v  cos 
 b   b  u 2   a   b   b   2   a 
Out-of-plane displacement, w, is approximated by a cubic polynomial:
 w1 
 3x 2 2 x 3   2x x 2   3x 2 2x 3   x 2 x   θ1   mπy 
w = 1 − 2 + 3  x1 − + 2   2 − 3  x  2 −    sin 
 b b   b b   b b  b b  w 2   a 
 θ 2 
The strain in the strip is decomposed into two portions: membrane and bending.
The membrane strains, εm, are at the mid-line of the strip and are governed by
plane stress assumptions. The bending strains, εb, follow Kirchoff thin plate
theory and are zero at the mid-line and a function of w alone.
ε = εm+ εb
 εx   ∂u ∂x 
   
εm =  ε y  =  ∂v ∂y  = N uv d uv = B m d uv
'

γ   
 xy  m ∂u ∂y + ∂v ∂x 

 εx  − z ∂ 2 w ∂x 2 
   
εb =  ε y  =  − z∂ 2 w ∂y 2  = N 'w' d wθ = B b d wθ
γ   
 xy  b 2z∂ w ∂x∂y 
2

As shown above εm and εb may be written in terms of appropriate derivatives of

the shape functions, N, and the nodal displacements, d. The elastic stiffness may
be understood through a statement of internal strain energy:
U= 1
2 ∫ σ εdV = 12 ∫ ε EεdV = 12 d ∫ B EBdVd = 12 d k e d
T T T T T

Where the stress is connected to the strain by an orthotropic plane stress

constitutive relation: σ = Eε, and E=ET. Since the membrane behavior (u, v) is
uncoupled from the bending behavior (w) separate elastic stiffness matrices may
be derived for each, where
k 0 
k e =  em
 0 k eb 

kem = ∫ B mT EB m dV and keb = ∫ B bT EB b dV

Substitution and integration leads to the following closed-formed solution for

the membrane, kem, and bending, keb, elastic stiffness matrices:
  aE1 abk 2m G  
  +  symmetric 
  2b 6  
 ak ν E ak G   abk 2m E 2 aG  
 m x 2 − m 
4 4   6
+ 
2b 

  
k em = t 
 aE1 abk 2 G   ak m ν x E 2 ak m G   aE1 abk 2m G 
 − + m  − −   +  
  2b 12   4 4   2b 6  
 
 ak m ν x E 2 + ak m G   abk 2m E 2 aG   ak m ν x E 2 ak m G   abk 2m E 2 aG 
 −  − +   + 

 4 4   12 2b   4 4   6 2 b 
where: mπ Ex Ey
km = E1 = E2 =
a 1− νxνy 1− νxνy
   13ab 4 12a 2  
 kmDy + k m D xy  
  70 5b  symmetric 
 6a 2 6a  
 + k m D 1 + D x  
5b b3 
 
  3a a   ab 3 4 4ab 2  
k m D1 + k m D xy 
2 2
 kmDy + k m D xy 
  5 5   210 15  
  2    
  + 3a D x + 11ab k 4m D y   +
2ab 2
k m D1 +
2a
Dx  
  b 2
420   15 b  
k eb = 
 9ab 4 12a 2   13ab 2 4 a   13ab 4 12a 2 
 kmDy − k m D xy   k m D y − k 2m D xy   kmDy + k m D xy  
  140 5b   840 5   70 5b  
 6a 2 6a   a 2 3a   6a 2 6a  
 − k m D1 − 3 D x   − k m D1 − 2 D x   + k m D1 + 3 D x  
 5b b   10 b   5b b  
 13ab 4 2
a   3ab 4 3
ab   11ab 42
a   ab 3 4 4ab 2 
 − k m D y + k 2m D xy  − k m D y − k 2m D xy  − k m D y − k 2m D xy   k mDy + k m D xy 
 840 5   840 15   420 5   210 15 
 a 2 3a   ab a   3a 2 3a   2ab 2 2a 
 + k D
m 1 + D x   − k 2m D1 + D x   − k m D1 − 2 D x   + k m D1 + D x 
 10 b2   30 b   5 b   15 b 
where: mπ Ex t3 Eyt3 Gt 3 ν yEx t3 νxEyt3
km = Dx = Dy = D xy = D1 =
a (
12 1 − ν x ν y ) (
12 1 − ν x ν y ) 12 (
12 1 − ν x ν y ) = 12(1 − ν ν )
x y

Geometric stiffness matrices

Consider the member to be loaded with linearly varying edge tractions (T1, T2)
as shown in Figure 1. The geometric stiffness matrix may be determined by
considering the additional potential energy incurred as these edge tractions
displace longitudinally (in the y direction); or equivalently in terms of higher-
order strain definitions (i.e., the Green-Lagrange strain):

x  1   ∂u   ∂v   ∂w  
2 2 2
a b

W = ∫ ∫  T1 − (T1 − T2 )     +   +   dxdy
0 0 b  2   ∂y   ∂y   ∂y  
 
The derivatives of the displacement fields may be written in terms of derivatives
of the shape functions, N, and nodal displacements, d, similar to the elastic
stiffness solution. For example, for bending, w:
2
 ∂w  T T
  = d wθ T N 'w N 'w d wθ = d wθ T G b G b d wθ
 ∂y 
Introducing this notation into the earlier potential energy statement leads to a
formal definition of the geometric stiffness matrix, kg.
1 T a b x 1
W= d ∫ ∫  T1 − (T1 − T2 ) G T Gdxdyd = d T k g d
2 0 0 b 2
Similar to the elastic stiffness matrix, kg may be broken into two parts. One
related to the u and v displacement fields, or membrane deformations, kgm, and
one related to w, or bending, kgb:
 ab
 x T
k gm = ∫ ∫  T1 − (T1 − T2 ) G m G m dxdy
0 0 b
ab
 x T
k gb = ∫ ∫  T1 − (T1 − T2 ) G b G b dxdy
0 0 b
Substitution and subsequent integration lead to the following closed-form
expressions for the geometric stiffness matrices:
70(3T1 + T2 ) 0 70(T1 + T2 ) 0 
 70(3T1 + T2 ) 0 70(T1 + T2 ) 
k gm = C
 70(T1 + 3T2 ) 0 
 
 symmetric 70(T1 + 3T2 )

8(30T1 + 9T2 ) 2b(15T1 + 7T2 ) 54(T1 + T2 ) − 2b(7T1 + 6T2 ) 

 b (5T1 + 3T2 ) 2b(6T1 + 7T2 )
2
− 3b 2 (T1 + T2 ) 
k gb =C 
 24(3T1 + 10T2 ) − 2b(7T1 + 15T2 )
 
 symmetric b 2 (3T1 + 5T2 ) 
where: C = b(mπ ) 1680a
2

Global stiffness matrices and assembly

To form the global elastic (Ke) and geometric stiffness (Kg) matrices each
individual strip must be transformed from local to global coordinates and then
assembled. A note on notation, local DOF are denoted by lowercase u, v, w, θ
and global DOF by uppercase U, V, W, Θ; further, stiffness matrices in the strip
local coordinate system are a lowercase k, and in the global system an uppercase
K is used. The local to global transformation at node “i” for strip “j” which is
oriented in the left-handed coordinate system of Figure 1 at an angle α(j) is
governed by the following transformations:
 u i   cos α ( j) sin α ( j)   U i   v i  1 0  Vi 
w  =  ( j)    and θ  =   
 i  − sin α cos α   i   i  0 1  Θ i 
( j)
W

These may be collected into a matrix form for all DOF in strip j as
d ( j ) = Γ ( j) D ( j )
Transformation of the stiffness matrices of strip j follows from:
T T
K e( j) = Γ ( j) k e( j) Γ ( j) and K (gj) = Γ ( j) k (gj) Γ ( j)
With all DOF in global coordinates the global stiffness matrices may be
assembled as an appropriate summation of the strip stiffness matrices, where:
j=1 to n j=1 to n
K e = ∑ K e( j) and K g = ∑ K (gj)
assembly assembly

Stability solution

For a given distribution of edge tractions on a member the geometric stiffness

scales linearly, this leads to the stability eigenvalue problem of interest, which
for a single eigen (buckling) mode, φ, and eigen (buckling) value λ may be
expressed as:
Keφ = λKgφ
Both Ke and Kg are a function of the strip length, a. Therefore, the elastic
buckling value and the corresponding buckling modes are also a function of a.
The problem can be solved for several lengths, a, and thus a complete picture of
the elastic buckling value and modes can be determined. The minima of such a
curve can generally be considered as the critical buckling loads and modes for a
member. CUFSM follows this implementation and an example is provided in
Figure 2. Given the selected longitudinal shape functions, a is generally known
as the half-wavelength.

0.9 SSMA 600S200-68

0.8
P y =40.11kips
0.7 Dist. P crd/P y =0.75

0.6
P cr / P y

0.5

0.4 Flexural
Local P crl/P y =0.42
0.3

0.2

0.1

0
0 1 2 3
10 10 10 10
half-wavelength (in.)
Figure 2 Conventional FSM analysis of an SSMA 600S200-68 structural stud
In the conventional FSM described above the selected shape functions result in
members that are pinned and free-to-warp at their ends. More complicated
boundary conditions are possible, but require the longitudinal deformations of u,
v, w to be defined in terms of a series (or splines) and result in much larger and
more complicated eigenvalue problems (see e.g., Bradford and Azhari 1995).
Further, more exact treatments of the deformations may employ higher order
polynomials for the transverse displacements (e.g., see Rhodes 2002). For a
complete treatment of finite strip derivations, and applications beyond member
stability, see Cheung and Tham (1998). Finite strip solutions that use the shape
functions presented here are the most common in thin-walled stability and will
be termed the semi-analytical or conventional FSM. CUFSM (Schafer 2006)
provides a solution in this form as does CFS (CFS 2006) and THIN-WALL
(Papangelis and Hancock 1995, 2006).

Definition of the buckling classes

As the FSM analysis of Figure 2 demonstrates, stability of cold-formed steel

members can typically be categorized into one of three classes: global (G),
distortional (D), or local (L). A convenient means for this classification is the
minima of the conventional FSM analysis results. However, while convenient,
this definition is by no means general and depends on the details of the cross-
section and loading. Sometimes minima may not exist, or extra minima may
exist. Qualitative definitions for the classes are also possible, for example the
Commentary to the Direct Strength Method (Appendix 1 NAS 2004), but again
such classifications are not general. Another characterization with some
popularity relies on the relative contribution from membrane versus bending
effects. For example, the strain energy, broken into terms derived from the
membrane (kem) and from bending (keb) are given in Figure 3 (as provided in
CUFSM). As the plots illustrate, each mode class has a certain characterizing
energy distribution over the cross-section: e.g., local modes involve only
bending strain energy. Definitive classification using any of the above methods
remains elusive.

Local Distortional Global

SE bending SE membrane

Figure 3 Strain energy related to membrane and bending deformations for the buckling modes
To provide a means of rigorous classification mechanical definitions have been
selected. Given the lack of alternative proposals, the criteria applied in
generalized beam theory (GBT) are used, (Silvestre and Camotim 2002a,b),
which are found to usually be in good agreement with current engineering
classifications. The separation of G, D, L and other (O) deformation modes are
completed through selective implementation of the following three criteria.
Criterion #1: (a) (γxy)m = 0, i.e. there is no in-plane shear, (b) (εx)m = 0, i.e. there
is no in-plane transverse strain, and (c) v is linear in x within a flat part (i.e.
between any two fold locations). Criterion #2: (a) v ≠ 0, i.e. the warping
displacement is not constantly equal to zero along the whole cross-section, and
(b) the cross-section is in transverse equilibrium. Criterion #3: κxx = 0, i.e. there
is no transverse flexure.

Table 1 Mode classification criterion

G D L O
modes modes modes modes
Criterion #1 – Vlasov’s hypothesis Yes Yes Yes No
Criterion #2 – Longitudinal warping Yes Yes No -
Criterion #3 – Undistorted section Yes No - -

Application of the criterion to the G, D, L, and O buckling mode classes is

defined in Table 1. Criterion 1 may be understood as being tied to classical
beam theory, or, Vlasov’s hypothesis, and restricts certain membrane
deformations, while allowing warping. Criterion 2 implies that warping
(longitudinal, or v deformation) must be non-zero, thus providing a separation
between local plate deformations which have no deformation at the mid-plane of
the plate and other deformation modes. Criterion 3 relates to distortion of the
cross-section and provides a means to separate G from D. While the modes
implied by these criteria are in accordance with current practice for a wide range
of practical problems; cases do exist where the applied definition leads to results
not in-line with current engineering practice. Further comparison of cFSM with
GBT is provided in Ádány et al. (2006).

Constrained Finite Strip Method

Application of buckling classes through constraints
Consider the membrane deformations duv of a single finite strip, as shown in
Figure 4. Introducing the first constraint from criterion #1, (εx)m=0 results in

(ε x )m = ∂u = −u1 + u 2 sin mπy = 0

∂x b a
and since the sine function is generally not equal to zero, u1 and u2 must be equal
to each other in order to satisfy the equality. This implies that the transverse
displacements of the strip’s two nodal lines must be identical, which is a natural
consequence of the zero transverse strain assumption. In practice, the identical u
displacements prevent those deformations where the two longitudinal edges of
the strip are not parallel, as illustrated in Figure 4.

Figure 4. Effect of (εx)m = 0 strain constraint on membrane deformations

The above derivation demonstrates that the introduction of a strain constraint

reduces the number of DOF, in this particular case from 4 to 3. Thus, we can
define the new, reduced DOF by u, v1 and v2, while the relationship of the
original and reduced displacement vectors can be expressed as follows:

 u1  1 0 0
 v  0 u
 1 =  1 0  
v1 or duv = Rduv-r
u 2  1 0 0  
    v 2 
 v 2  0 0 1  
where R is the constraint matrix, which is a representation of the introduced
strain constraints. For more general strain-displacement constraints (i.e., the
other criteria), and more general cross-sections, the derivations are more
complicated, but finally the associated constraint matrices (R) can be defined, as
shown in Ádány and Schafer (2006a,b) for G and D modes and Ádány (2004)
for L and O modes, and thus apply for all the criterion summarized in Table 1.
Since a different R matrix may be constructed for each of the modal classes: G,
D, L, and O, taken together they span the entire original nodal FSM basis and
represent a transformation of the solution from the original nodal basis to a basis
where G, D, L, and O deformation fields are segregated, so, for any vector of
nodal displacements in global coordinates, D:
D = [RG RD RL RO]Dr = RDr
represents the transformation. The partitions of the R constraint matrices are the
deformation fields associated with the G, D, L, and O spaces that meet the
criterion of Table 1, while columns of R corresponds to individual deformation
modes For the SSMA 600S200-068, R for the G, D, and L spaces are provided
graphically in Figure 5. For example, the first four deformations in Figure 5a
and b provide the warping displacements and transverse displacements for the G
modes, i.e.
RG = [G1 G2 G3 G4]

G1/O1 G2 G3 G4 D1 D2
(a) warping displacements of RG and RD

G1/O1 G2 G3 G4 D1 D2
(b) transverse displacements of RG and RD

L1 L2 L3 L4 L5 L6 L7 L8

L9 L10 L11 L12 L13 L14 L15 L16

(c) transverse displacements of RL

L1 L2 L3 L4 L5 L6 L7 L8
(d) transverse displacements of first 8 RLu modes when transformed to unit-member axial modes
Figure 5 Deformation modes spanning G, D, L for SSMA600S200-068 based on cFSM

An important characteristic of the RG and RD deformations is that the transverse

displacements are uniquely defined by the warping displacements. For RL
Figure 5c provides the transverse displacements (note, no warping occurs in the
L modes). As shown, these L modes appear to be in the original FSM nodal
DOF basis, but they can be transformed into a modal basis, as discussed below.

Additional transformations inside the G, D, L, O spaces are possible and

desirable. One attractive option is to use unit-member axial modes as detailed in
Ádány and Schafer (2006b), which, as the name suggests, are the modes for a
member of unit-length under axial load within a given space, for example RL is
transformed to RLu as Figure 5d demonstrates. These modes may have
physically desirable characteristics in some situations. Finally, the O modes
(RO), associated with shear and transverse extension are not shown here, but are
discussed in detail in Ádány (2004) and Schafer and Ádány (2006).

Modal decomposition

In the conventional FSM the complete stability problem is solved. However, in

cFSM it is possible to constrain the deformations down to only a specific class
and thus decompose the modes. For example, consider an analysis focusing only
on distortional buckling. Constraint of the deformations from the nodal
coordinates to the generalized coordinates in the D space is accomplished by:
D = R DD D
Application of this transformation on the eigen stability problem results in a
constrained eigen stability problem:
RDTKeRDφD = λRDTKgRDφD or KeDφD = λKgDφD
This decomposed stability problem is much smaller than the original stability
problem. For the SSMA 600S200-068 example used in this paper the original
FSM model consists of 40 DOF, but the decomposed model consists of only 2
DOF – the generalized deformations associated with D1 and D2. While the
model reduction is potentially beneficial, even better is the reduction in
mechanics – restriction to a selected space (e.g., D) allows one to study a mode
in detailed ways that cannot be completed in conventional FSM analysis.

The G, D, and L modes are decomposed from the general solution using the
procedure described above and the calculated critical forces are plotted against a
conventional FSM solution for the 600S200-068 section in Figure 6. The L and
mode solution shows excellent agreement with the conventional FSM minima.
The D mode solution suggests a somewhat stiffer response than the conventional
FSM. For distortional buckling, bending in the web is greater in a conventional
FSM than in the decomposed cFSM solution (this is shown in the insets to
Figure 6, but also may readily be seen in strain energy plots). The relative
difference between the conventional FSM minima for D and the cFSM solution
is typically smaller in most lipped channels studied by the authors; but in
general a difference exists. (Note, this implies a difference between GBT
distortional mode solutions and conventional FSM solutions as well). The G
mode curve shows the same tendency as a conventional FSM solution, for long
members, however the critical loads are somewhat higher. The difference is due
to the fact that in-plane transverse deformations are fully restricted in G modes,
as discussed in Ádány and Schafer (2006b).
1.5
FSM
L
D
G

600S200-068
Pcr/P y

D by FSM D by
0.5 minima cFSM

0
0 1 2 3
10 10 10 10
half-wavelength (in.)

Figure 6 FSM and cFSM solutions for SSMA 600S200-068

Modal identification

It is desirable to understand how the different buckling classes (G, D, L, and O)

contribute in a conventional FSM stability solution. Indeed, one long-term goal
of this research is to provide a classification method that may be used not only
in FSM but in general purpose FEM as well. Any displaced shape, D, or
buckling mode, φ, may be transformed into the basis spanned by the buckling
classes through the use of R, via
Dr = R-1D or φr = R-1φ
where the coefficients in Dr (or φr) provide the contribution to a given column of
R, or when summed over columns the contribution in a given class. However,
the coefficients are dependent on the normalization of the columns of R. A
definitive normalization scheme has not been finalized. If Ri is the ith column
vector of R, two normalization schemes are considered here.
Vector norm: normalize by setting ||Ri|| = 1. Any displacement is expressed as a
linear combination: φ = ΣciRi = Rc, participation, pi, of mode i is defined by pi =
|ci|/Σ|ci| where the summation is over all modes.

Weighted strain energy norm: normalize by setting ½RiTKeRi = 1. Any

displacement is expressed as a linear combination: φ = ΣciRi = Rc, participation,
pi, of mode i is defined by pi=|ci/λi0.5|/Σ|ci/λi0.5| where the coefficients are
weighted by the eigenvalue in that mode alone, i.e. λi = RiTKeRi/RiTKgRi.

The contribution of a mode class which spans n columns of R is defined as:

n modes
∑ pi n
i =1

Using these definitions (implemented in CUFSM) modal identifications are

performed on the SSMAS200-068 and reported in Figure 7. Both methods show
the expected regimes of L, D, and G; however it is also clear that the different
norms do provide different notions of the response. For instance, at the half-
wavelength of the distortional minima L has a non-negligible participation with
D. This provides further commentary to the results of Figure 6 where a D only
solution for stability proves stiffer than a conventional FSM solution.
1.5

1
Pcr /Py

0.5

0
0 1 2 3
10 10 10 10
100

50 L G
D
vector norm
0
0 1 2 3
10 10 10 10
100

G O
50 L
D
weighted strain energy
0
0 1 2 3
10 10 10 10
half-wavelength (in.)
Figure 7 Modal identification results for SSMA 600S200-068

The vector norm is perhaps the simplest possible participation scheme, and it is
most similar to that used by GBT; however it lacks a physical basis, is
dependent on the discretization of the member, is dependent on the system of
units the problem is solved in, and generally discounts the impact of rotations as
translations are typically several orders of magnitude larger. The weighted strain
energy norm provides a physically motivated normalization and reasonable
identification results. The O mode contribution observed in Figure 7 is
consistent with the findings in Schafer and Ádány (2006) which demonstrates
that conventional FSM includes an O mode contribution in global buckling.

Discussion

As shown in Figure 6 and Figure 7 minima identified in a conventional FSM

solution are not necessarily pure mode solutions. This has ramifications for
design methods that employ these critical buckling values. In addition, it further
complicates the application of simple qualitative modal identification methods
and suggests that quantitative methods such as that provided in cFSM are
needed. Further discussion of challenges related to cFSM is provided in Schafer
and Ádány (2006).

Conclusion

The conventional finite strip method combined with the constrained finite strip
method provide a powerful tool for exploring cross-section stability in cold-
formed steel members. In the conventional finite strip method elastic and
geometric stiffness matrices are formed from a summation of cross-section strips
and employed in an eigenvalue stability analysis. The stiffness matrices are
explicitly derived in this paper and can readily be used in engineering software.
The provided solution is identical to that employed in the open source stability
analysis program: CUFSM. The constrained finite strip method is described and
examples of its application provided. The strength of this new extension to finite
strip solutions is the ability to decompose and identify conventional finite strip
solutions as related to buckling classes of interest: global, distortional, or local
buckling. The examples provided here show the potential use of the constrained
finite strip method, and the algorithms discussed are implemented in CUFSM.

Acknowledgments

The presented research has been performed with the financial support of the
Korányi Imre Scholarship of the Thomas Cholnoky Foundation, the OTKA
K62970 of the Hungarian Scientific Research Fund, the American Iron and Steel
Institute, and CMS-0448707 of the United States National Science Foundation.
References

Ádány, S., Schafer, B.W. (2004). “Buckling mode classification of members with open
thin-walled cross-sections.” Fourth Int’l Conf. on Coupled Instabilities in Metal
Structures, Rome, Italy, 27-29 Sept., 2004
Ádány, S., Schafer, B.W. (2006a). “Buckling mode decomposition of single-branched
open cross-section members via finite strip method: derivation.” Elsevier, Thin-
walled Structures, (In Press)
Ádány, S., Schafer, B.W. (2006b). “Buckling mode decomposition of single-branched
open cross-section members via finite strip method: application and examples.”
Elsevier, Thin-walled Structures, (In Press)
Ádány, S., Slivestre, N., Schafer, B., Camotim, D. (2006). “Buckling analysis of
unbranched thin-walled members: generalized beam theory and constrained finite
strip method.” III European Conference on Computational Mechanics, Solids,
Structures and Coupled Problems in Engineering, C.A. Mota Soares et.al. (eds.),
Lisbon, Portugal, 5–8 June 2006
Bradford, M.A., Azhari, M. (1995). "Buckling of plates with different end conditions
using the finite strip method." Computers and Structures, 56 (1) 75-83.
CFS (2006). CFS Version 5.0, RSG Software, www.rsgsoftware.com, visited on April
25, 2006.
Cheung, Y.K., Tham, L.G. (1998). The Finite Strip Method. CRC Press.
NAS (2004). 2004 Supplement to the North American Specification for the Design of
Cold-Formed Steel Structures. American Iron and Steel Institute, Washington, DC.
Papangelis, J.P., Hancock, G.J. (1995). “Computer analysis of thin-walled structural
members.” Computers & Structures, Pergamon, 56(1)157-176.
Papangelis, J.P., Hancock, G.J. (2005). Thin-Wall: Cross-Section Analysis and Finite
Strip Analysis of Thin-Walled Structures, Thin-Wall v2.1, Centre for Advanced
Structural Engineering, University of Sydney,
http://www.civil.usyd.edu.au/case/thinwall visited on 15 March 2005.
Rhodes, J. (2002). “Post-buckling analysis of light gauge members using finite strips.”
Sixteenth International Specialty Conference on Cold-Formed Steel Structures.
Orlando, FL USA, October 7-18, 2002.
Schafer, B.W., Ádány, S. (2005). “Understanding and classifying local, distortional and
global buckling in open thin-walled members.” Tech. Session and Mtg., Structural
Stability Research Council. Montreal, Canada.
Schafer, B.W., Ádány, S. (2006). “Modal decomposition for thin-walled member stability
using the finite strip method.” SMCD 2006, May 14-17, 2006, University of
Waterloo, Waterloo, Ontario, Canada
Silvestre, N., Camotim, D. (2002a). “First-order generalised beam theory for arbitrary
orthotropic materials.” Thin-Walled Structures, Elsevier, 40 (9) 755-789.
Silvestre, N., Camotim, D. (2002b). “Second-order generalised beam theory for arbitrary
orthotropic materials.” Thin-Walled Structures, Elsevier, 40 (9) 791-820.